EP 514: Google’s AI Studio - 5 time-consuming tasks you didn’t know you can automate
01:12:48 Share

This is the Everyday AI Show, the everyday podcast where we simplify AI and bring its power to your fingertips. Listen daily for practical advice to boost your career, business, and everyday life. You probably know a lot of Google's AI products and offerings, right? Maybe you use Google Gemini.

As a chat bot, like you would chat GPT. Maybe you use Google Gemini inside your workspace apps like Google docs and Google sheets, or maybe you use one of my favorite tools from Google powered by Google Gemini, NOPA Galem. But there is a legit hidden gem that I think most people don't know about.

And that is Google's AI Studio. And I think for a lot of use cases, it actually might be Google's best AI. And I don't think many people know and most people aren't using it. And there's been a lot of recent updates just in the last few weeks that I think make Google AI Studio even better.

All right. So we're going to be going over that today and talking about inside Google AI Studio five time consuming tasks that you probably didn't know you could automate.

All right. I'm excited for this one. I hope you are too. If you're new here, what's going on y'all? My name is Jordan Wilson and I'm the host of Everyday AI. We're a daily live stream podcast and free daily newsletter helping us all not just learn AI, but how we can leverage it to grow our companies and grow our careers. Is that what you're trying to do? If so,

It starts here. This is where you learn with the live stream, with the podcast, but where you leverage it and where you actually put it to work, that's on our website. Number one, on our free email newsletter that we put out every single day, we're going to be recapping the highlights

of today's show. So maybe you're driving or walking your dog and you're like, oh, what was that? It's going to be in the newsletter. But also while you're there, we have more than 500 episodes. We've interviewed hundreds of the world's leading experts on AI across a variety of topics. So no matter what you're trying to learn, our website is where you go and make that happen. It is a free generative AI university.

All right. Normally, most days we go over the AI news. There's actually a ton going on, but I didn't want this one to turn into too long of a show. Wanted to keep it very factual, you know, hard hitting. So if you want the AI news, that's going to be in our newsletter. All right. Here we go. Let's just jump into. Well, no, I got to shout out all my people here. McDonald joining from Chicago. Yeah, we're doing this live, y'all. I like to say.

you know, at everyday AI, we're the realest thing in artificial intelligence, right? So much of what you see online or listen to or watch, right? It's, it's very fabricated. It's very scripted. That's not what we're going to do here. We're going to be trying to do a lot of this stuff live. There's going to be some bugs. I'm sure. Uh, so thanks for joining us. Uh, McDonald, uh, Kelly joining from Syracuse, uh, Dennis joining on LinkedIn from New York city. Love to see it. Uh, Noel joining from Chester, England. Uh, uh,

Aiden, not too far away from Indiana. Jackie, happy Tuesday. I'll be seeing Jackie tomorrow at DePaul. That'll be fun. A lot of people. Got to shout out Dr. Harvey Castro joining us from Berlin, Germany. All right. Let's get into the use cases, y'all. And it is Tuesday. So here's my hot take for Tuesday.

Google AI Studio is the best AI tool you've probably never heard of, right? If you're an avid AI user, right? You're using large language models hours a day. You've probably heard of AI Studio, but if you're more casual, you probably haven't, or you probably at least don't know what it's capable of. I'm curious, live stream audience. We've got a ton of people in the house today. How often do you use AI Studio?

or have you never used it? I want to know. I'm also kind of curious because I think most people either haven't heard of it or they don't know what it's capable of. And we're going to be going over those five time-consuming tasks that we can do, that we can automate. So let's start at the top. What the heck is Google AI Studio? Well, number one,

It looks a lot cleaner than it did a month ago, right? So a shout out to the Google team. They really cleaned, uh, cleaned up the, uh, user interface and the user experience. And well, why am I mentioning the user interface and user experience, uh, for an AI tool? Well, it is more for developers, right? So, uh,

at least initially, Google AI Studio was really made for developers. It's a place where people who are a little more technical can go in, they can work on some of their code, they can work on structured prompts, they can work on how they're using Google's API on the backend. So let's say that your company is maybe fine tuning one of Google's models or

You're using their new, you know, Gemma 3, which I think is an amazing small model, right? So maybe you're using that, you know, on device, on-prem, you're fine tuning it, whatever it may be. But, you know, I think Google AI Studio was originally just created to make developers' lives easier, right? In the same way that OpenAI has their playground. But the difference is AI Studio is like legit a full platform.

fledged working product. And it's for much more than developers. I actually think the everyday non-technical person such as myself, I'm not super technical, right? Um, I'm not fine tuning models aside from just testing it out inside of AI studio. But I think for non-technical everyday people, there's so much to explore so much. You probably don't know, uh,

in Google AI Studio. So yes, originally kind of set up for AI developers over the last, I'd say six months, it's really grown into so much more. And I think that there's a lot of tools and features that a lot of other companies have kind of promised, but maybe haven't fully delivered on yet. And Google AI Studio has delivered. So is anyone using this? All right, so Allison says a handful of times.

Heather says, heard of, haven't used. Angie says, barely. All right. Kelly hasn't used it. Jay hasn't used it. All right. A lot of people haven't. Noel says, just heard of it. Dennis is using it weekly. All right. So some of you are using it. Some of you have never heard of it. So buckle up. We're going to go into it. But at its core, originally kind of made to help developers, you

make better apps with Google's API, right? Because there's so many tools and services out there, tens of thousands that you use that are powered by either Google Gemini on the backend, OpenAI, Claude, et cetera, right? So developers usually have kind of like a sandbox or a playground that they can go experiment and make their products better before releasing it to us all in the world, right? So think of those

you know, all these AI tools and you're like, wow, how do they work? Well, generally they're running on, you know, Google open AI or Claude, um, and developers need a spot to go and play around with it and make it better. And that's why, you know, talking about the UI UX, there's a lot more, right? So if you just log on to Gemini.google.com, it's an AI chat bot, right? You're just chatting with Google Gemini. You can still select your model, but for the most part, you don't have a lot of controls.

Inside Google's AI studio, there's a lot of controls. So you can control the temperature, change your model, toggle certain tools on or off for structured output, code execution, function calling to bring in kind of third-party APIs, grounding with Google search, which is new and amazing.

amazing my gosh it is actually it got the uh the the top spot on the lm arena board for internet connected search inside google ai studio it's so good okay so that's kind of a the the gist of it and here's the other thing it's free yeah it's free uh some more updates here but let me put out uh kind of uh

Let me read first what Google says about this kind of unpaid services versus paid services, because there's an update here that I think people are going to like, because previously, you know, Google AI studio, it just anything that you put in there, it used that information to train its models.

right? So, and that is still the case for a free. So if you are on a free Gemini account, this is what Google says and how they use your data. So it says when you use unpaid services, including for example, Google AI studio and the unpaid quota on, on Gemini API, Google uses the content you submit to the services in any generated responses to provide, improve and develop Google products and services and machine learning technologies. All right. So

If you are on a free Google plan, you probably shouldn't be uploading confidential proprietary data from your company because Google, if you are on the free plan, it does use that to train its model inside AI Studio. However, and this was just updated, I believe in April. So just like

This month, uh, this was updated. So now if you are paying for Google Gemini and you have a cloud billing account set up, not technically, you know, it's not hard. Took me like 30 seconds when I originally set up, uh, you know, this, uh, I don't know, a year or so ago, but here's what it says now. So it says now when you use paid services, uh,

including for example, the paid quota of the Gemini API. Google doesn't use your prompt or responses to improve our products. So.

That's an update. I don't even know if Google really announced this and I didn't really see anyone talking about this, but y'all, I always do my homework before I do these shows. I don't just randomly jump into something. I care about data privacy, security, how, you know, hallucinations, truthfulness, how the model works, right? So good thing Google updated this. So if you're on a paid plan and you set up that cloud billing, Google's AI studio does not use anything that you input to train its model. If you're on the free plan, it does. But y'all, let me be honest.

So much of the information that you're probably going to put into any chatbot, it's already on the internet. It's already public information, right? Let's just put that out there. All right. A lot of people don't understand how much of, you know, your company's data is already in large language models. If it's been posted on the internet ever before, it's in the models. All right. So if you are brand new and feel free live stream audience to follow along as we go, the first time you go into Google's AI studio, you're going to get an option.

Okay. And just, just so you know, it is AI studio.google.com. Okay. So it's going to say it's time to build and you can either use Google AI studio and click that try Gemini. So yes, it is Gemini inside of AI studio, or you can do a build with Gemini API. So if you do want to, you know, you can actually use Google's API key, which a lot of people didn't know you could do. So you can, you know, try it out that way.

All right. A couple housekeeping things to keep in mind. One, if you go up to settings, you know, you can do light mode, dark mode. But yeah, like I said, the user interface used to be a lot more geared toward highly technical people. Now it's easy to follow. It's so much cleaner now. But you're going to want to turn on the auto save option. So go up to your settings, toggle that auto save on. Otherwise, by default, Google's AI studio does not save your chats. Okay.

Now let's go into a use case. Number one, long context window. All right. Here's where we're going to get a little tricky and apologies y'all for probably jumping around a lot of a lot of tabs here. So what, what I have is I went through

And I'm going to share this with our live stream audience podcast audience. This is one of those you might you might want to watch the video on this. So if you're listening on the podcast, we always leave a link. You can go watch this video, whether it's on YouTube, LinkedIn, whatever. OK, so I pulled about 50 episodes.

All right. So from our AI news that matters. All right. It is nearly a 400 page documents. All right. So I'm going to copy this right here. All right. It might literally crash my computer just from copying so much information. So why does long copy?

context window matter and why is that my number one use case for Google's AI Studio? Well, you'll see here, even with OpenAI's new O3 model, it can't handle this much context, right? So I'm telling it to at the very top of this prompt,

I am saying, please carefully analyze this content and find 10 underlying trends that indicate trends from AI companies, large language models, generative AI, et cetera. So again, I have the transcript of our AI news that matters shows. I think there's probably like 50 episode transcripts. So that's about 40 hours of transcripts content. So if I go into, uh,

Open AI's new model, their O3 model. I love it. I think it's one of the most capable models in the world. I think Gemini 2.5 Pro is probably a little more useful right now, but I think O3 is probably more capable. And I did full shows on each of those in the last two weeks, but I can't even put this full context in here, even with the extended context window of ChatGPT's O3 model.

right? You might be saying, oh, okay, Jordan, what about Claude? Claude has a long context model. Can't handle this. Can't handle this. So you'll see from a context, this is 250,000 tokens. All right. So more than a quarter million tokens. It's a lot of information, right? On nearly 400 pages of transcripts. So Claude,

can't handle it it says it's 40 over the length length limit so i can't do anything uh even google gemini right on their front end because this is one of those reasons you might be thinking okay well why would i ever use ai studio can't i just use all the features and functionality inside google's front end chatbot at gemini.google.com well no here's one use case already i can't paste in all that information uh into uh into google gemini it just cuts it off but

but I can obviously in Google's AI studio. All right, so let's see if I can do this live. If not, I do have a screenshot, but let's see. All right, so I am going to copy and paste literally 400 images

400 pages worth of context. So now I am in Google AI Studio. I'm not going to go over the entire interface now because we're going to be doing some of these things later. So I'm going to choose, just so it's a little faster, I'm going to choose Gemini 2.5 Flash Preview, which was just released about 12 days ago. So you do have on the right hand side, you have these different models that you can choose from.

And the good thing I love about this inside Google's AI studio, it gives you, so if you are a developer and you want to care, oh, how much might this cost if I put something like this out into production? For each model, you can just hover and it gives you, you know, thinking input tokens, non-thinking input tokens.

What it's best for, use cases, the knowledge cutoff. I wish every single model had this. All right, so I'm just going to go to Gemini 2.5 Flash Preview. I'm literally just pasting in 450.

pages of content. All right, I'm going to scroll to the top. So again, all I'm saying here at the top is please carefully analyze this content and find 10 underlying trends that indicate trends from AI companies, large language models, generative AI, et cetera. So I want to know essentially over the last eight months,

everything that's been happening in the world of AI news. What are the trends? I might know some of these things. I might not. And I am using a Gemini 2.5 flash preview, which is a thinking model. But I did use the flash version, which is the smaller version versus the pro because I'm guessing the pro one would have taken longer in my gosh, this is done. That was like no time at all. 26 seconds. It went through and it used reasoning. My gosh.

If you're listening on the podcast, this is one of those ones where I'm just like smiling and I'm like, I can't believe this technology exists. Number one, I can't believe this technology is free. Insane. My gosh. Okay. So, and it did find 10,000,

different trends. I'm just going to read just the headings, but it gave me details, right? And the crazy thing is people are always like, oh, Jordan, how do you remember so much on AI? Well, number one, I forget a lot, right? I forgot most of this stuff, even though it came from my lip.

right? But so many of these things were back in like eight months ago. So I forget things. I use Google AI Studio every single day to remember and recall information. So let's just for fun, go over the 10 trends. Number one, escalating competition in rapid model responses. And it's some very good details there. Number two, the rise of agentic AI and task automation. Again, we're looking at October 2024 through April 2025. So

Yeah, back in October, there really wasn't a lot of talk on agentic AI or even task automation inside large language models. Now there is. Number three, advancement in specialization in multimodal AI. Four, heavy investment in innovation in AI hardware and infrastructure.

Five, diversification and experimentation with pricing and business models. That's a good one. Talking about, as an example, OpenAI's pro tiers. Anthropic just came out with some new tiers. So yeah, there's details supporting all of this as well. Six, increasing strategic partnerships, investments, and acquisitions. Seven, growing focus on AI safety, ethics, and governance. Eight,

rising regulatory scrutiny and geopolitical tensions. That's a huge one, right? And to be able to know just over the course of five to six months, here's what's happening there. That's huge. Nine, transformation of user experience and accessibility. And then 10, significant impact on the workforce and industry structure. This is so good. And because I used a thinking model, I can go in inside Google AI Studio and look and see

how it went through all of this information. So extremely impressive. I can go see how the model thinks, see how it planned and see how it identified these 10 different trends. And yes, it would have been better if I use Gemini 2.5 Pro, but it probably would have taken another minute or two. And I'm trying to get through these different use cases a little quicker. All right.

Are you still running in circles trying to figure out how to actually grow your business with AI? Maybe your company has been tinkering with large language models for a year or more, but can't really get traction to find ROI on Gen AI. Hey, this is Jordan Wilson, host of this very podcast.

Companies like Adobe, Microsoft, and NVIDIA have partnered with us because they trust our expertise in educating the masses around generative AI to get ahead. And some of the most innovative companies in the country hire us to help with their AI strategy and to train hundreds of their employees on how to use Gen AI. So whether you're looking for chat GPT training for thousands,

or just need help building your front-end AI strategy, you can partner with us too, just like some of the biggest companies in the world do. Go to youreverydayai.com slash partner to get in contact with our team, or you can just click on the partner section of our website. We'll help you stop running in those AI circles and help get your team ahead and build a straight path to ROI on Gen AI. Yeah, Angie just says, whoa. Yeah, I agree.

Curtis from YouTube says those little model cards are so helpful. Yes, I agree. It's I wish number one, I wish that that was also in the front end of Google Gemini, but I wish everyone did that, right? Those model cards are so, so helpful, right? Especially if you're a beginner, right? And you go down to Gemini 2.0, right? Because there's actually some features that are only available in Gemini 2.0 inside Google AI studio, not in the newer Gemini 2.0.

You have your Gemma models. So extremely, I think probably still the world's best small language model in Gemma 3. But then you could also go and play with some of these preview models, which is the LearnLM models.

All right. So that's number one. You'll see already. That's something that no other platform could take. A million tokens, right? In the front end. It just... That's nutty. So think of all the use cases. And I'll tell you what I do for this all the time. So yes, that's...

you can do something very similar in notebook lm uh right you know dump a bunch of contacts in there uh and this could fit inside notebook lm but one of the reasons is i still like sometimes having flexibility um outside of your grounded information so as an example and let me just do this quick all right so uh

All right, we're still on use case number one, but there is this new thing grounding with AI search. So now I can use information that's not in that 400 pages, right? Whereas on Google's notebook LM as powerful as it is, and it literally won our 2024 AI tool of the year. But now I can click this new option grounding with Google search, right? And I can say something like, you know, based on April, 2025 movements,

in the AI space, please bullet point what's new for the above 10 trends. So like as an example, if I just want to see what's happened this month with those 10 trends and to see where they're at in this current day, because some of them are reflecting back six months or more, I can ground this in Google search and then I can go see exactly what it's doing. So now it's just going to get information from

April, 2025 only. And it can go out and search and use the web. So very cool. All right. Use case number two, video creation. All right. I'm going to try to do this, try to do this live. Let's see how it goes. All right. So

People don't know you can create videos inside of Google AI studio. So one thing I'm technically now starting inside of Gemini advance because I want to get a photo first. So I'm going to, unfortunately, I don't think right now inside AI studio, you can create images, but you can create videos with their new VO two model. So I am in Gemini right now and I'm saying create an overhead

overhead aerial photo, realistic shot on full frame DSLR from above a commercial airplane flying over the Chicago skyline. All right. So now I, again, I am in the front end of Gemini. I don't necessarily like this. This isn't what I was going for. It only, it only got the, the wing there. I wanted to be able to see the entire,

the entire plane. So I just clicked the redo. I already did this once. It was actually a really good photo if I'm being honest. But I wanted to be able to see the full airplane overhead. If I can't get exactly what I'm looking for, I did run this once this morning and I'll use that image, but I was trying to see if I could get a nice one. All right. It did the same thing, but that's fine. So anyways, you know, you can pull this into Google AI Studio. All

All right, so now I'm jumping back into Google AI Studio and I'm going to go on the left-hand side. You have this new video gen. All right, so...

vo2 is so good right uh i'd still say even though we have some new updates as an example runway gen 4 we've had some very impressive offerings uh from some of the uh chinese companies uh cling uh as as an example all the different ai video companies i still think vo2 is ahead it's ahead of sora it's ahead of everyone else all right so uh google

FYI, Google is bringing VO2 if you have a paid version of Google Gemini. And it's a slow rollout because you'll see right now on my different, I have four different paid Google accounts. I don't have VO2 access inside of Google Gemini right now. So the Google Gemini chatbot, gemini.google.com. But if I go into Google AI Studio, I do have it. So on the left-hand side there, you'll see I have VideoGen.

Okay. And the good thing is inside Google's AI studio, there's a ton more options. So on the right hand side, you have all these different settings. So I can create a 16 by nine video. I can create a vertical video, nine by 16, you know, anywhere from five seconds to eight seconds. I can change the frame rate. I can have a set the resolution or run a negative prompt, which is if something keeps popping up in my AI video generation and I don't want it to, I can write it in the negative prompt box.

But the good thing with Google's AI studio and using VO two there versus, yeah, you can use it in the more enterprise vertex and get even more options. But the thing I like inside Google AI studio is you can start with an image because unfortunately,

If you're using this inside Google Gemini, the last I checked, you aren't able to start with an image. So this is great. So I uploaded the previous image that I had and I have a little prompt here. So I'll read it out.

So all I'm saying is cinematic panning shot showing the airplane slowly moving from right to left across the Chicago skyline. All right. And here is the image that I ultimately used. So I don't know. Live stream bodies. Does this look does this look like a real image? Right. Google's imagine this.

Google's Imagine3 AI photo generator is really, really good, right? If you look at like LM Arena, which, you know, we might as well just do that now because it's going to be a second for this to generate. But if you look at LM Arena and if you go to the leaderboards and you go to, let's see, text to image. So this is the different...

text to image. Like imagine three from Google is the top AI image generator, at least that they have access to inside this program. All right. So let's see here. Hey, this wasn't too bad. Michael said video generating on a live demo. Jordan is brave. Yeah, I know. Right. All right. But it looks like it's already done. All

All right. So I'm going to go ahead, put this full screen. So we have an eight second video here, live stream audience. Let me know how does this look again? So we started with the image in Google Gemini, brought it over to Google AI studio using the video gen on the left hand side inside Google, Google AI studio. And this is all for free. Oh, this is wild. This is wild. All right, here we go. Here's the photo or the video.

Not bad. Not bad. So pretty good. There were some artifacts right here. There's some little blur here. But again, you can just regenerate. And the good thing is, it's just using natural language. I just said cinematic panning shot showing the airplane slowly moving from right to left across the Chicago skyline. So what could you use this for? I mean, my gosh, you're

companies marketing is probably still stuck in the 1990s. It's old, it's dry, it's dull. So let's say as an example, your big annual conference this year is in Chicago, right? If it is in Chicago, by the way, reach out to me. I'll come keynote your conference and run a workshop. But maybe you're trying to put together a nice looking landing page for

for your annual conference, for your big company's annual conference in Chicago, this would be a great, I think, you know, get this like three second video and put that as a background on the website, right? It's great. You can tell if you're from Chicago, you can tell it's Chicago, right? For our podcast audience, it looks really cool. I mean, to be able to get a shot like this, right? You would have to be in another area

airplane because you can see the entire airplane flying over the Chicago skyline. So very cool, right? Is this going to be in the next Netflix documentary? No. But again, I essentially just two-shotted this, right? I took two attempts at it. It's really good. It's really good. Yeah, livestream audience,

What do you think of this video? I think it's like in terms of being like putting up on your website as a background video, you know, putting it on social media, right? You could throw it in Canva, something like that. Add some effects, add some words in terms of you now have this used to take so much time, right? To get a video like this.

five years ago would be nearly impossible and extremely expensive. Now, in terms of your marketing, your comms, your advertising, using Google's VO2, especially if it's just going to end up like as a smaller something on your website, social media, et cetera. Again, I don't think you would want to use this at least right now, right? In a feature film or something like that. But

It looks extremely realistic, especially on a smaller screen. If you're only doing a couple of seconds, it's very good. And where this takes your creativity, like next level, next level. Yeah. Angie just says, I can't believe it's free. Jay says, actually at an annual, oh, Jay, I suck. I know you texted me. I mean, I got to get back to you. All right. So

Sandra says, I'm actually doing it on my computer as we're speaking. And it's super amazing. Yeah, it's, it's so, so good. I think the physics and the realism, specifically the physics is what separates VO2 from Google ahead of Sora. Sora is great, but it struggles with physics. I think there's some creative tools inside Sora from OpenAI that I really like, but it struggles with physics. It struggles with realism. Google VO2 crushes it. It is absolutely

absolutely so good. Yeah. Denny says good for background image that might have an overlay. Exactly. I would put this if I was having a big conference in Chicago, right? And you're selling tickets or whatever it is, you know, you're trying to bring thousands of employees together for whatever, right? Or just, you know, maybe your website needs a refresh. Go do it like this, right? It's really good.

All right. So that is use case number two. Let's go to use case number three. I've done this before, but parsing PDFs. Okay. And I know this might sound like a boring one and not super impressive. All right. But

But it is because, again, up until the most recent line of models from both Google and from OpenAI, this was something that large language models really struggled with. Right. You know, using a combination of OCR. All right. Yeah.

So that is OCR is optical character recognition. So, you know, these models would use a combination of OCR and computer vision and it wouldn't always work. All right. So let's go again. We're doing this live. I did do this one other time, but I want to include this on today's show because I think it's extremely impressive. All right. So let's jump over. We're doing this live. We are in Google's AI studio. All right.

There we go. And all I'm going to say is transcribe this PDF word for word. All right. And for this one, uh, I am going to go with Gemini 2.5 pro the most powerful model. All right. I'm going to click run and we'll see how long this takes. And first of all, you might be thinking, okay, why, why might I want to use this? All right. How many times have you gotten a super long PDF? Um,

and maybe you have to make updates or you're like my gosh i have to read through this entire thing uh and it's it's it's gonna take forever and maybe i just need a little bit of information out of here uh so in this case and a lot of times different tools struggle reading pdfs right especially uh you know i'll show you all an example of of what this pdf is going through and you'll see why this could be extremely helpful right so this is going through

This is kind of our sponsorship opportunities deck. So if you look through this, it's images, right? Any tool would struggle to read this. I created this in Canva. There's a bunch of images. A lot of this stuff that looks like text is

isn't even text, right? It's literally a bunch of images. Any program would struggle. So when companies reach out and they want to advertise with us, I send them this little deck over. So I want to pull up this page. So at the bottom, it says trusted by leaders from, right? So we have hundreds of

of subscribers from big companies like Google, Amazon, Nvidia, Microsoft, IBM, Intel, et cetera. So we have their logos at the bottom of this deck to tell people, yeah, if you wanna reach our audience, we have a lot of very reputable people who are building the future of AI reading our platform. All right, so that's a logo, okay? So now it's done, all right? So let's see how long that took.

Didn't take long. So again, you can go in if you're using these thinking models, these reasoning models, you can go through. It took 53 seconds. All right. And here's the thing. It got it done.

Perfectly, perfectly. It got all those stats. It pulled all this information. It pulled all this information from, you know, images that would be very hard for any OCR or computer vision to do, right? So in a lot of these pages, it's just like 10 images and maybe a little bit of text. So right here, it even pulled the logos. These logos are not perfect.

labeled. Let me repeat that. These logos are not labeled. So as an example, the Amazon logo, it's just an A. It's just an A with the little smiley face, right? If you know Amazon's logo, you probably do. It literally knows that that's Amazon's logo, even though the word Amazon isn't on there. So do you see the utility and how this can be a huge time saver? So maybe you have to

update a PDF and you're like, oh my gosh, who has that original file? I'm going to have to redo all of this. Maybe you need to read information from a PDF and it's a long, you know, 100 page, you know, industry white paper, but you only care about 10% of this. Well, Google AI studio is a great place to do that because when you combine this very large context window with the fact that it is one of the smartest and most capable models in the world, I mean, you have a tool here that is almost unbeatable.

unfair. Uh, right. And I'm not even going to get into, into the more technical aspects of what you can do. Uh, but you can, even if you wanted to, you could create like a version of Google gems, right. Or, you know, if you've used custom GPTs inside of chat GPT, so I could save

Let's take that industry white paper example. Let's say every single week, there's a huge industry PDF that comes out. It may be a big part of your job is reading through it, pulling out some of the key insights and then briefing your team. And maybe only 10% of it pertains to you, right? That's probably something a lot of you do, but you could essentially go in here. You could create essentially a version of a Google gem. It wouldn't be called that, but you can put in custom instructions.

You can also go in here if you wanted to get a little more technical, you could even turn on the structured output. So maybe, hey, in each week, there's always a table or a chart

And maybe there's eight columns, but you only want three of the columns because that's all that matters. You could go in. It's very easy, right? You could essentially give examples to Google AI Studio and then create that. And then, you know, it would save because you can save it with a system prompt and then you can toggle the option on for Google.

structured output right so i can go in here i can go into this visual editor and i can add a property so you know uh and then you can export this as what's called the json all right uh i know that sounds a little technical but essentially you can you know kind of uh save different types of prompts and train it on input outputs but you can just do it in natural language all right

What do you guys think about use case two? Richard says, how about had that been written text in a PDF? It would have been done much faster. Much faster. Yeah.

Curtis says it'll even accurately extract text from a PDF that has wonky OCR text in it. Yes, it is so good. Allison says I pay for Acrobat Pro and still receive this error every time I have it use OCR if there's a single rasterized image. Yes, exactly. It says, oh yeah, Acrobat cannot run OCR as the page contains renderable text.

and then she says ai's video just did this successfully in five seconds uh yeah it's it's it's nutty y'all this is so good are you guys again this is free the fact this is free and we're only halfway through our use cases that was number three all right number four multi media content creation okay so this is one of those

I'm gonna go ahead and jump over into my screen. Again, podcast audience, we're doing this live if you wanna see. All right, so for this example, I'm actually going to use an older model, all right? And I do think and hope that they're going to update which models have this capability. So for right now, I'm gonna go to my model selector. I'm gonna go to Gemini 2.0.

and then i'm going to go to this flash image generation all right so there's actually a lot of things and i wasn't going to do this but i saw someone in the comments had asked for it um so i'm just going to do an example of this this isn't even one of my use cases uh but let me see here i'm gonna try to give me a second here all right i'm gonna try to upload uh an image uh let's see give me a second y'all i know this is uh

Doing these things live, never fun, never fun. All right. Let me, okay. So I'm going to pull in

this same photo that we used. Okay. So this new, uh, you know, and it has the word hot on it, right? If you hover over, uh, this is essentially, uh, you know, we've, we've been hearing a lot about the GPT four Oh image gen amazing. Uh, right. So, uh, Gemini 2.0 flash, the experimental version, and it has image generation. This isn't my use case, but someone asked for it. So I'm going to see if I can do it here live on the fly. Uh, so I, I,

uploaded that same Google Gemini image of the plane flying over the Chicago skyline. So let's just say, I'm going to try this. I'm going to say, keep everything else the same and turn the, and I'm going to say, but make the plane bright yellow with the words,

Jordan air on the side. All right. I don't know if this is where, if this will work, hopefully it does, but you know, I've been mentioning on this show with between what it's done already. All right. It didn't do a great job. It actually added an additional plane. Uh, so I, I probably should have done, uh, I probably should have done this a little better. I'm going to say, uh, there should be only one plate. All right. We'll see if that, uh, if that works. Um,

So I've been talking about on the show between GPT-4.0 image gen, between this new thing from Gemini 2.0 Flash, like everything that Canva announced. I think there's going to be a lot of pressure on Adobe because this is like, if you get it right, this is like Photoshop. This saves so much time. All right. So there we go. It did it correctly. It changed the perspective a little bit.

It made the plane slightly larger. So let's even say that took five seconds. I'm going to say, uh, you know, the plane is too large, make it more like the original image. Uh, keep the one plane yellow with Jordan air on the side. All right. If I was taking a little bit more time, I think I could get this really well, but y'all, I can't tell you, right. I've been using Photoshop since, uh, how old am I? I'm a little old. Uh,

I've been using Photoshop for 20, 23 years, almost a quarter century. I've literally wasted thousands of hours that now you can get a pretty good job done inside Google's AI studio. All right. So that one didn't work perfectly. It added some fragments of another image. That one was a little complicated, but go in and play with that. But that's not even the use case. All right. So I got distracted. So let's get back to the use case, which in this case, it is.

doing multimedia content creation. So again, I'm inside Google's AI studio. On the right-hand side, I chose Gemini 2.0 Flash. And you'll see the big thing here is it has this image and text output. All right. So all I'm going to say is, you know, write a blog post on the top five tourist spots in Chicago and create photos for each of them. All right. That probably just to type that

FYI, probably took me 12 seconds. Let's see how quickly Google Gemini can create this. So, all right. It's at, it's, let's see, is it done yet? Not yet. Okay. So you'll see, it's literally writing me a blog post, but it's giving me the top five must-see spots in Chicago. But for all of those five in line, it's giving me

It is using Imagine, which I'm actually curious if this is using Imagine 2 or Imagine 3. My hunch is it's using Imagine 2 because I've run these exact same prompts inside Google Gemini with the Imagine 3 and in the Imagine 3 interface. And the quality is a little better. So I'm not sure. I'll try to get that verified. But I believe if you're using this, I think it's using Imagine 2. But regardless,

I mean, y'all look at these photos, right? The fact that I just quote unquote in 23 seconds, right? I have a blog post. Let's say you're, you're, you're a tourism company and you have a new tour here in Chicago with the five, you know, the five, uh, you know, hottest spots in Chicago for tourism. The blog post is written. The images are there, right? Are these the most high quality images? Absolutely not.

Absolutely not. But they're really good. Right? Hey, another secret here. And what do you think? What do you guys think of this? This right here? I don't think any other large language model, any popular one by default anyways, can do this, right? Literally, I just in theory, quote unquote, wrote, I would want to go through human in the loop, spend more time on the front end, spend more time on the back end. But I have a blog post here with five photos.

that Google Gemini generated. I used to do a lot of SEO content creation and I've been doing that on the web probably since, I don't know. I mean, I built my first website in the 90s. I was paid to build some of the first websites I did in what, like 2010? So I spent a good three, four, five years mainly doing this, blog posts with images for a lot of clients. It just did it in 20 some seconds.

And it's really, really good. All right, here's a little cheat code, right? Let's say you accidentally go into the wrong mode. So let's say you go into Gemini 2.5 Pro and you run this exact same chat, okay? It's obviously not going to output images, right?

But what's really cool because it's a thinking model and it's like, yo, I can't create images. You'll see this because I always like to test this. I'm like, okay, what happens if you ask a model inside a Google AI studio to do something that it is not capable of? And I can obviously read its kind of chain of thought to see how it's processing this. Let's see if it did it. Here we go.

So what it does is it's like, I can't generate an image, but it gives you a very detailed prompt. And then I can go into...

and then I can go into Google Gemini and just run that prompt, right? So I just said, create an image for, and I copy and pasted the image description that it gave me. So Google Gemini 2.5 Pro realized, hey, I can't create an image even though you asked me to. So instead, here's an image description that you could put inside. And yeah, this one, again, this photo is much better

than the one that was generated inside the multimodal version inside Gemini 2 Flash. So I do think or I assume it's a better image generation model. All right. We got more, y'all. But I made a mistake. So I made a mistake.

There's actually an extra one that I wanted to do that I think is better than the rest of these. I know there's still a couple between all our different platforms. There's more than 100 people that are still around. So if you want me to get to the sixth one, just say the word bonus. All right.

Sometimes I like to have fun, see who's still here paying attention. So if you want me to do the sixth one, because actually the sixth one, initially I said, oh, here's five time-consuming tasks. And then this morning when I was going over my show, I'm like, wait,

I didn't put the newest and most powerful one in this list. So if you want me to do the sixth one, just type the word bonus. If we get 10 people, I'll do it. If not, maybe I'll share about in the newsletter. All right. So just put in the word bonus. If I get 10 people. All right. So here we go for number five, this one might get a little tricky. So I'm sorry. Also a live stream audience aside from typing in the word bonus, if you want to see this, um, I'm going to need some help.

on this because I want you all to hear this one because number five is learning any new skill via share screen stream, okay? And this is one of those. This is one of those that, you know, OpenAI previewed this almost a year ago and we still don't have it on desktop. But right now, Google Gemini, again, for free can see your screen

And you can interact with it. Let me repeat that for free. Google Gemini's AI studio can see your screen and you can interact with it. All right.

Yes or no. Be honest, y'all. How many of you knew that? How many of you all knew that? All right. So I'm going to get something going here. I hope that you can hear it. I'm going to have to move my mic. I've done some testing with this. Even when I share, even when I share the audio, you can't always hear it. So number one, live stream audience, can you still hear me? All right. Because I moved my mic away.

All right. And then I'm going to go ahead and stop sharing this. And I'm going to go over, I'm going to go over this use case, which I think is amazing. All right. So give me a second, y'all. Doing this live, it's dangerous. It's dangerous here. Okay. So let me first describe what I have. Okay. I have a simple spreadsheet.

from my Google Search Council data. And of the 10 pages with the most clicks, I exported that and put it into a Google Sheet. Okay. And what I want to know is,

I want to know the difference in impressions over the course of the last seven days. All right. So did this go up 10%? Did it go down 30% in terms of impressions? So these are some stats that I'm pulling in. And so essentially what I want to do is I need to write a formula. Okay. And as much as I love Google Sheets and Excel and everything else, does anyone struggle with formulas?

Just me. Anyone else? All right. So now I'm going to go into Google AI Studio. All right. So that is the sheet I am going to share. All right.

And hopefully you can hear this. So you might not be able to. I apologize. I'm cranking my volume all the way up. So here we go. So on the left-hand side of the screen, I'm clicking the stream icon. So this, I can just talk to Gemini live. I can show it my webcam so it can see me or if I was trying to explain something because this also works on mobile, which is really, really cool. Or I can click share.

share screen so this is what i'm going to try to do and again this is something that uh open e i demoed uh they i think they of 2024 and we still don't have it on desktop all right so i'm going to click share screen hopefully you can hear so i'm clicking share screen and then i have to choose uh i have to choose which one we're going to do uh hey gemini can you uh hear me

Yes, I can hear you. How can I help you today? Cool. Can you tell me what's on my screen? On your screen, I see a Google Sheets spreadsheet titled...

EDAI last seven. The spreadsheet contains data organized in columns such as top pages, clicks last seven days, impressions last seven days, and difference. Do you want me to describe any of the data in more detail? No. So I understand the data. So in this column here, H2, I want to write a formula that shows me the percentage change in this impressions difference column.

between the column E2 and F2. So G2 gives me the number, but I want to turn that into a percentage. And I'm not sure the formula that I need to write in H2. So can you tell me the formula? To calculate the percentage change in cell H2, you can use the formula equals G2E2.

After entering the formula, format the cell as a percentage to display the result as a percentage value. Will that work for you? All right. Yeah, cool. Give me a second. So I'm going to highlight. So you just gave me this formula here inside the Google AI Studio chat. So I'm going to copy that. Give me a second here. And I'm going to go over. I'm going to insert it into the formula bar. Click on it.

click enter and there we go. Okay, cool. So now it gave me like a fraction. So that's fine, but I want it to show up as a percentage. How can I change this new column, the H column to a percentage?

I understand that you want to convert the values in the H column to percentages. To do that, select the entire column by clicking on the column header. Then, in the toolbar, find the format menu. Within that menu, there should be an option for number or format as. Choose the percent option from the drop-down menu. This should format all the numbers in column H as percentages. Does that make sense? Cool. All right. Did I do that right? Can you look? Did I get it correctly?

Yes, the H column now displays the values as percentages. You did it correctly. Is there anything else I can help you with?

Yeah, maybe just out of curiosity, and if you could be just very direct and blunt and act as kind of an SEO strategist. So, you know, these are all website pages from my website and some Google Search Council data. What should be some of the first things if I want to increase traffic to my website by looking at this screen? Where do you think my best opportunity is or what should I go do right now?

From the data, it appears that the web page Free Chat GPT vs. Chat GPT Plus What's the Difference is experiencing a significant decrease in impressions and clicks. To increase traffic, you should investigate why that page is experiencing such a large drop. Could you tell me about any changes you've made to this page recently? All right. So now what I'm doing is I stopped that. What do you guys think? Was that pretty impressive?

All right. Angie just said my clone. Um, and the good thing is, which I didn't have this enabled. I could have done grounding with Google search. Right. Uh, and then it could have planned that episode, uh, for me. Right. Which is really, really cool. So do you guys understand?

the magnitude of that. I just gave a simple example, right? Okay. Uh, I would have probably, you know, that, that, that was a very easy, uh, you know, formula to write. Like, let me be honest. It wouldn't take me that long. Um, but maybe if I was trying to do more, something more advanced probably would have taken me, I don't know,

5, 10, 15, 20 minutes to go find something, right? When I can give Google's AI Studio and Google Gemini access to my screen, I can become an expert in anything, especially with this new option that no one's talking about that you can ground with Google search. Because at the end, it correctly identified that one of my articles on my website,

which brings in the majority of the traffic, it's losing ground. And the reason why is, well, that article is a little old. So it's probably about six months old. And there's a lot of things in the new chat GPT free versus paid that have changed. So it's losing ground. So if I would have had grounding enabled, I could have talked with Gemini to help me plan that

update, it could have looked at that webpage, said, okay, here's some new updates that have happened with ChatGPT that are not in your article yet. So all of a sudden, if you use this correctly, think of it. Think of like you now have the smartest person in the world over your shoulder. You can ask them anything.

and they're going to get it. So, you know, you might be trying to learn a new piece of software, uh, right. Maybe it's something like I showed you an Excel sheet. Maybe you're coding something. Maybe you just want feedback on a design. Uh, right. And it's actually really good. So you do need to do a little bit of prompting. Uh, sometimes the, uh, the live talking models are a little verbose. So I always like to tell it to be direct, um, you know, to never be wishy-washy or politically correct. Always, you know, give me hard hitting, uh,

and actionable advice because sometimes, and that's all large language models, they might give you a bunch of fluff. So you can still do a little bit of best practice prompt engineering and really kind of, you know, give it a role, give it a task, you know, have a conversation with it, iterate, evaluate, et cetera, and you're going to get great results. But you can literally learn anything, anything new that you're trying to learn

You can do it. This, I don't know why more people aren't talking about this and using it, especially the fact that you can now combine it with this grounding with Google search. Y'all, y'all. Angie says, I'm going to spend all day doing this. McDonald's said, I did not know that.

Fred, good question. Fred says, can it make the changes you requested to the spreadsheet without you having to do them? So I don't think so. Although Fred, strangely enough, I ran this simulation a couple of times. Now, one time Google Gemini did say, hey, do you want me to put that formula in there? And I said, yeah, sure. It didn't do it. But, you know, maybe. Monica said, I fought with a spreadsheet for hours last week. So yeah, maybe now you can go fight less.

Big Bogey says, how is this free? Yeah. Good, good stuff. So yeah. All right. Let's see. Did enough people want the bonus one or should I end this? Let me see. How many people said bonus? One, two, three, four, five. I don't know. Maybe. Let's see. All right. Well, at least let me stop sharing this screen. Take a drink. We're going to go. Angie. Angie says next prime, prime polish with Gemini. Maybe.

All right. I think I think we got 10. I was trying to count. I won't be I won't be a strickler, but OK, yeah, we got OK. Now that I'm scrolling up, it looks like the word bonus showed up way more than 10 times. All right. I know it's already been an hour, but let's let's do it. Here's number six in this one. Y'all is even worse.

crazier. All right. I am not kidding. All right. Let me get my, uh, gosh, where's my, uh, link there. There it is. Okay. All right. So give me a second. Number six is better YouTube learning. Stick with me here. It's more than you think. All right. Uh, so what I'm going to do. All right. So I had this video. Uh, when was this video? Give me a second.

Okay. This video was from seven months ago. So it's a little old. It is a nine minute video and it's going over. Oh gosh. Looking, looking at my hair on this video, I need a haircut right now. It looks like I just got a haircut there. So,

It's a nine minute video and the video is Google's notebook L M the AI tool you can't ignore. So I think this was my second video that I did on a notebook LM. Okay. So all I'm doing here is I'm copying the URL. This is a nine minute video. Okay. So now.

I have my screen sharing going on here. Okay, so this is new. This is new. So at least using it in the Gemini 2.5 Pro. So one thing I didn't talk about is there is this little plus button. If you go over in the middle by the run button,

Okay. So if I click the plus button inside Google AI studio, there's different options where I can start with different media. So the other amazing thing about Google AI studio, it's multi modal by default, but it can ingest video.

but not in the way that you think. And let me show you that and let me prove that. So again, when you think of this large context window and the fact that Google's AI studio can technically see this, let me not exaggerate here.

Actually, I'm not going to say anything. I want you all to see it live first. So I can click when I click that plus button, I can click my drive, upload file, record audio, camera, YouTube video, sample media. Okay. So all I'm going to do here is I'm going to click YouTube. I hope this works. And then a box pops up. It says YouTube video URL. I'm going to paste it in there. It loads the thumbnail, which these thumbnails are silly, but someone from Nvidia told me that my video

My YouTube was lame and I had to make it better. So there's a thumbnail, I guess. So all I'm going to say is, what do we want to say here? Give me a second. So I'm just going to type in. So this is a video about Notebook LM. So I'm going to say bullet point the steps on how to use Notebook LM.

Okay. And then I'm going to click run and you'll see right here. It gives me a token count. So this nine minute video is 157,000 tokens. So I'm going to click run prompt. Okay. And again, I'm using Gemini 2.5 pro. So presumably this is going to take a little longer. Uh, all right. So, uh,

Jackie says Google owns YouTube. So this makes sense. Will it work with other video channels? Great question as we wait, Jackie. So yes, you can also just upload a file if you have a video file. Now, as we wait for this and I'm going to go ahead and click this and you can see in the chain of thought, Gemini 2.5 Pro is thinking it's going through, it's crunching this video and it's almost already done. Okay.

This is good. Okay. So yeah, it's already done. So it took 42 seconds and it gave me bullet points on what I talked about in this video. So number one, access notebook LM, uh, go to the notebook LM interface, create a new notebook, add your sources, process the source, locate audio overview, um, generate the conversation. Okay. So pretty cool. So it just broke down a nine minute video, uh, into bullet points. All right.

And I probably get what you're saying. Like, okay, Jordan, what's the big deal? Why is this changing how we work? Okay. Let me, let me do something else. So I'm typing, describe the person in the video. Hmm. Because there's so many AI tools that you can throw in a YouTube link. And you know, what it essentially does is it goes through the transcript. Okay. And then it's like, Hey, according to the transcript, uh, here's, here's what, um,

Here's, oh gosh, hey, at least this. Okay, whew, all right, that's good. So y'all look at this. Based on the video, here's a description of the person. Appearance, he is an adult male appearing to be Caucasian, likely in his 30s or early 40s. Ouch, Google, ouch. Hair, let's see.

All right. Hair. He has dark brown or black hair. My wife will love that one. We always argue about my hair color. I like to pretend that my hair is brown, but it's probably black. Hey, and at least Google was a little nicer here. And it didn't mention all of my gray hair. But facial features, clean shaven, attire. Look at this.

He is wearing a dark color, likely black or dark Navy short sleeve polo shirt. The shirt has thin white stripes along the edge of the collar. A small light colored logo appears to be CK visible on the left side of his chest. Y'all look at this.

You can barely even see that, right? So apparently I have a, looks like a Calvin Klein polo right there. That little CK is tiny, but do you understand and realize what is going on here? Okay. Think of how this can change the work that you do.

Let's say, I don't know, for an easy example. Let's say you work in a factory, okay? And whatever it is you process, let's say you make these mugs, okay? I have a glass mug here from Nespresso. Like Nespresso, I've name dropped you like 50 times. Why do I not yet have a lifetime subscription to Nespresso, right? Anyways.

Let's say your company makes these Nespresso mugs, right? And you have video that is monitoring. And let's just say it's a 30 minute video, all right? And a lot of things go wrong during this 30 minutes, right? You're doing a tour, some things knock off. One of the pieces of the machine malfunctions, right? A couple of the glasses come out wrong, whatever. I'm trying to give you an easy example on something that I saw. Google Gemini can go through and do all of this,

and say, hey, how many glasses fell off the conveyor belt? Give me the time and then give me a classification on what caused it to fail, right? And then remember, you can have these structured outputs as well as system prompts. So let's say during that 30 minutes, 100 glasses fell off and 1,000 went through successfully. So it was a 10% failure rate. So you could go in there

create a system prompt using structured outputs where it essentially kicks out

you know, a spreadsheet looking output that gives you the timestamp, what happens, and it can classify it as well. Was it human error? Was it machine error? Was it, you know, faulty glass, whatever. You can create those classifications. Google's AI studio can go in and essentially watch 30 minutes of, you know, let's say it's security video. You upload it there. My gosh,

Livestream audience is no one's, uh, is no one's mind blown anyone else. You know, we just have some people saying this is crazy. Um, y'all think of the new possibilities and just how this changes the future of work. Right. Um,

A lot of people, if you are in a role where you're observing something all day and you have to report back, you have to classify this information. Y'all Google AI studio does that, right? Can you upload an eight hour file right now? Probably not, but it's getting, it's getting there, right? So it's even describing my demeanor in this video.

It says he appears engaged and expressive as he talks, looking directly toward the camera slash viewer. He smiles and shows enthusiasm while discussing the topic. My gosh. Like the fact that this technology is available today, it's available for free and it can completely change the way you do business. Period.

That's a wrap y'all. All right. Let me quickly recap our six time-consuming tasks that you didn't know you could automate or run inside Google AI studio. Number one, long context window. Number two, video creation. Number three, parsing information from PDFs. Number four, multimedia content creation. Number five, learning any new skill via screen reading.

share screen stream and our last bonus one better youtube learning and just being able uh to literally have a large language model be eyes for any video and to be able to classify information much more than a transcript it can see and use a brain it's wild that was a wrap y'all i hope this was helpful all right uh but there you have it if this was helpful

Please share this. I would appreciate if you share this. If you tag someone that needs to know this, I think Google's AI studio is getting absolutely slapped on. Like I said, a couple of these updates. I don't think what I just showed you, the ability to upload a YouTube video in Gemini 2.5 Pro. I was using prior models before this. I think that's brand new.

And again, it's not like the thousands of other AI tools that do that, that just look at the text, they look at the transcript, they look at what was said, they look at the words. No, this can see, it can use its brain and it can use the Gemini 2.5 Pro model. So if this was helpful, please let me know, repost this, share someone, tag someone that needs to know this in the comments, y'all. I'd like to make Everyday AI a place where we can learn together, where we can share, but you gotta do your part too. You gotta share, you gotta give me feedback.

on what's working and what's not. So I appreciate you listening. I know this was a long one. If you're on the podcast, thank you for sticking with us. Like I said, this might be one of those ones where you need to go click the show notes, go watch this video. It'll hopefully explain things a little better. So thank you for tuning in.

Go to youreverydayai.com. If you haven't already, sign up for the free daily newsletter. We're going to be recapping this one. It's not going to take you an hour plus to get through the newsletter. It'll be about seven minutes for you to become the smartest person in AI at your company. Thanks for tuning in. Hope to see you tomorrow in Every Day for more Everyday AI. Thanks, y'all.

And that's a wrap for today's edition of Everyday AI. Thanks for joining us. If you enjoyed this episode, please subscribe and leave us a rating. It helps keep us going. For a little more AI magic, visit youreverydayai.com and sign up to our daily newsletter so you don't get left behind. Go break some barriers and we'll see you next time.